Files that include <sys/reent.h> on i686-pc-cygwin can cause GCC 4 (10-26-04) to bomb out due to a segfault in the compiler itself. I don't have a backtrace (I don't know how to get one if the ICE occurs during the bootstrap process itself, which it does).
Created attachment 7832 [details] Preprocessed source of the header file in question This is a preprocessed copy of a file that did nothing else but include <sys/reent.h>.
10-26-04 is very old, I cannot reproduce it with a cross compiler with today's compiler, can you try a new snapshot.
This bug still exists with GCC 4 as of 10-27-04, and I miscategorized this bug. It occurs during bootstrap in builds hosted on i686-pc-cygwin.
Do you mean 12-27-04 or 10-27-04?
(In reply to comment #4) > Do you mean 12-27-04 or 10-27-04? Sorry, I meant 12-27-04. The 10 was a typo.
Ok, then I will close this as invalid to ...
Mark as unconfirmed.
(In reply to comment #7) > Mark as unconfirmed. Ok, does anyone else have Cygwin?
I finally figured out how to get a backtrace from the stage2 compiler, so here it is (this is from running the posted testcase thru cc1): --backtrace starts-- Starting program: /home/Lucas/b-gcc/gcc/cc1.exe /home/Lucas/test.i Program received signal SIGSEGV, Segmentation fault. 0x00583da6 in emit_insn (x=0x101d19a0) at ../../gcc/gcc/emit-rtl.c:4375 4375 rtx last = last_insn; (gdb) bt #0 0x00583da6 in emit_insn (x=0x101d19a0) at ../../gcc/gcc/emit-rtl.c:4375 #1 0x00646f9c in ix86_handle_cdecl_attribute (node=0x22e614, name=0x101d7a48, args=0x0, flags=0, no_add_attrs=0x22e61b "") at ../../gcc/gcc/config/i386/i386.c:1716 #2 0x0040965f in decl_attributes (node=0x22e6f0, attributes=0x101da870, flags=0) at ../../gcc/gcc/attribs.c:253 #3 0x00405d9e in yyparse () at c-parse.y:1723 #4 0x004087c2 in c_parse_file () at c-parse.y:2922 #5 0x0047bf62 in c_common_parse_file (set_yydebug=0) at ../../gcc/gcc/c-opts.c:1092 #6 0x0049ec9b in compile_file () at ../../gcc/gcc/toplev.c:992 #7 0x004a079e in do_compile () at ../../gcc/gcc/toplev.c:2085 #8 0x004a0801 in toplev_main (argc=2, argv=0x10100f08) at ../../gcc/gcc/toplev.c:2117 #9 0x004972e0 in main (argc=2, argv=0x10100f08) at ../../gcc/gcc/main.c:35 --backtrace ends--
(In reply to comment #9) > I finally figured out how to get a backtrace from the stage2 compiler, so here > it is (this is from running the posted testcase thru cc1): This backtrace does not makes sense as ix86_handle_cdecl_attribute does not call emit_insn, are you sure you don't have an extra patch which is applied wrong?
(In reply to comment #10) > (In reply to comment #9) > > I finally figured out how to get a backtrace from the stage2 compiler, so here > > it is (this is from running the posted testcase thru cc1): > > This backtrace does not makes sense as ix86_handle_cdecl_attribute does not call emit_insn, > are you sure you don't have an extra patch which is applied wrong? OK, I updated my CVS to 10-28-04 and I still get this error (same place, too: the libgcc2 build is what's failing). I'll try doing a fresh co.
Fresh co as of 10-28-04 works. It was a patching glitch. =P Thanks!